LLM API Cost Calculator
Prices verified · 2026-08-03 · 12 model tiers
1
Choose a development workload
3
View methodology Like-for-like model comparison
400K tokens/h · 80:20 · 25% cache
| Model | Input / 1M | Output / 1M | Cache | Hourly | Daily | Monthly | Source |
|---|---|---|---|---|---|---|---|
| Claude Sonnet 4.6Anthropic Current | $3 | $15 | $0.3 | ¥14 | ¥70 | ¥1,540 | Anthropic list prices ↗ |
| DeepSeek V4 FlashDeepSeek | $0.14 | $0.28 | $0.0028 | ¥0.4 | ¥2.02 | ¥45 | DeepSeek API ↗ |
| DeepSeek V4 ProDeepSeek | $0.435 | $0.87 | $0.0036 | ¥1.25 | ¥6.27 | ¥138 | DeepSeek API ↗ |
| Gemini 3.1 Flash-LiteGoogle | $0.25 | $1.5 | $0.025 | ¥1.31 | ¥6.55 | ¥144 | Gemini API ↗ |
| Qwen 3.7 PlusAlibaba Cloud | ¥2.998 | ¥11.991 | ¥0.6 | ¥1.73 | ¥8.63 | ¥190 | Alibaba Cloud Model Studio ↗ |
| Claude Haiku 4.5Anthropic | $1 | $5 | $0.1 | ¥4.67 | ¥23 | ¥513 | Anthropic list prices ↗ |
| GPT-5.6 LunaOpenAI | $1 | $6 | $0.1 | ¥5.24 | ¥26 | ¥577 | OpenAI API ↗ |
| Gemini 3.5 FlashGoogle | $1.5 | $9 | $0.15 | ¥7.86 | ¥39 | ¥865 | Gemini API ↗ |
| Gemini 3.1 ProGoogle | $2 | $12 | $0.2 | ¥10 | ¥52 | ¥1,153 | Gemini API ↗ |
| GPT-5.6 TerraOpenAI | $2.5 | $15 | $0.25 | ¥13 | ¥66 | ¥1,441 | OpenAI API ↗ |
| Claude Opus 4.7Anthropic | $5 | $25 | $0.5 | ¥23 | ¥117 | ¥2,566 | Anthropic list prices ↗ |
| GPT-5.6 SolOpenAI | $5 | $30 | $0.5 | ¥26 | ¥131 | ¥2,883 | OpenAI API ↗ |
Calculation method
Hourly cost = uncached input × input rate + cached input × cache rate + output × output rateAll token rates are per million; concurrent agents scale both input and output volume.
Official pricing sources
Prices last verified 2026-08-03. Open each provider source from the comparison table.
Billing boundaries and limits
Tools, code sandboxes, web search, and file retrieval may be billed separately.
Long-context tiers: Some models apply higher request-level rates beyond context thresholds.
Cache writes and storage: Explicit cache creation, storage, taxes, and channel charges are excluded.
Platform markups: Bedrock, Vertex AI, Azure, and resellers may differ from direct provider rates.
Scenario bands are editable planning assumptions, not industry averages or model-speed guarantees.